[Speex-dev] reg. speex Ie Values

Leandro Galvão lleandrogalvao at gmail.com
Mon May 2 10:21:56 PDT 2011


I'm using the ITU-T P.834 methodology to derive the Ie parameter for
Speex. According to the E-model, Ie indicates the degradation in
quality of a codec at 0% packet loss. It is a non-negative value: the
higher, the worst. Details in ITU-T G.107.

Because Speex is multi-rate, we are deriving one Ie parameter for each
Speex rate. In our preliminary experiments, we've got a value of Ie=0
for Speex at 24.6kbps, which is the same value for the G.711 (64kbps).
All other codecs has generally Ie above 0, because of lossy
compression. Thus, I'm amazed that Speex at 24.6kbps is giving us a very
good value of Ie=0. Looking at the PESQ x bitrate graph presented at
Speex home page (http://www.speex.org/comparison/), I saw that the
last red cross point of Speex is very close to the pink square point
of G.711, which corroborates to our findings. We've also used PESQ in
our experiments, so I'm afraid that our results can be only another
way of reading this graph.

So, I would like to hear from you guys if this result of Ie=0 for
Speex at 24.6kbps does make any sense?  There are any human-based
evaluation study of Speex that can be used for comparison?


Leandro Galvao

More information about the Speex-dev mailing list